TurboVGGT: Fast Visual Geometry Reconstruction with Adaptive Alternating Attention
NeurIPS 2026
Adaptive sparse attention for faster multi-view 3D reconstruction.

Computer Engineering · University of Toronto
I am a Computer Engineering undergraduate at the University of Toronto, graduating in 2027. My research spans video generation, 3D vision, and multimodal reasoning.
I aim to build models that can generate, understand, and reason about the visual world. My recent work covers efficient 3D reconstruction, controllable video generation, and reasoning in latent spaces. I am applying to PhD programs for Fall 2027, and I am also open to industry research roles.
–
Deep Learning Research Intern
Video diffusion models for image editing and facial reenactment.
Research Intern
Efficient 3D reconstruction, video generation, and open-vocabulary scene understanding.
Student Researcher
Efficient multimodal LLM reasoning with reinforcement learning.
* Equal contribution
NeurIPS 2026
Adaptive sparse attention for faster multi-view 3D reconstruction.
Preprint · 2026
Exploring multiple latent hypotheses with an RL controller that allocates reasoning compute.
CVPR 2026 Findings
Generating consistent RGB, depth, and semantic videos across multiple driving views.
Preprint · 2026
LangSVR unifies scene appearance, semantics, and geometry in sparse voxel representations.